Objective: We report an experiment in which three training approaches are compared with the goal of training adaptive teams. Background: Cross-training is an established method in which team members are trained with the goal of building shared knowledge. Perturbation training is a new method in which team interactions are constrained to provide new coordination experiences during task acquisition. These two approaches, and a more traditional procedural approach, are compared. Method: Assigned to three training conditions were 26 teams. Teams flew nine simulated uninhabited air vehicle missions; three were critical tests of the team's ability to adapt to novel situations. Team performance, response time to novel events, and shared knowledge were measured. Results: Perturbation-trained teams significantly outperformed teams in the other conditions in two out of three critical test missions. Cross-training resulted in significant increases in shared teamwork knowledge and highest mean performance in one critical test. Procedural training led to the least adaptive teams. Conclusion: Perturbation training allows teams to match coordination variability during training to demands for coordination variability during posttraining performance. Although cross-training has adaptive benefits, it is suggested that process-oriented approaches, such as perturbation training, can lead to more adaptive teams. Application: Perturbation training is amenable to simulation-based training, where perturbations provide interaction experiences that teams can transfer to novel, real-world situations.